Identification of methylation QTLs in breast cancer characterises the influence of germline SNP variation on the abnormal tumour methylome

Abstract

An important factor in expanding our knowledge of transcriptional regulation has been to better understand the role of epigenetic modifications in mediating transcriptional activity. One hallmark of this relationship is the observation that gene activity associates with DNA methylation patterns. Abnormal changes to the methylome are a pathological feature of many cellular disease and a hallmark of cancer. However, the mechanisms underpinning these alterations remain unclear. Current models in normal tissues have suggested that some methylation changes may occur in an allele-dependent manner whereby genetic variation associates with both local and global changes in the normal tissue methylome. Investigating the mechanisms and directionality of the interplay between DNA methylation, sequence variation and their role in gene regulation may therefore reveal important insights into tumorigenesis. Previous studies have established the effects of genetic variation on the methylome of normal tissue. However, a similar characterisation of the genetic effects on the cancer methylome has been limited due to the complexity of analysing tumour tissues which have highly heterogeneous cell content compositions.

I used a population genetics approach to examine the effects of genetic sequence variation on the methylome of breast tumours. I asked if there were loci which showed an allele-dependent association with the methylation state of CpGs (termed methQTLs) either locally in cis or distally in trans. I then asked if I could use the distinct mechanisms which underpin cis-methQTLs and trans-methQTLs to characterise the role these genetic variants may have in regulating the methylome. I developed a robust analytical framework for detecting methQTLs using breast tumour methylation data and matched normal blood germline genotype data. I compared and assessed various methods for the pre-processing, normalisation and quality control-based filtering of these datasets to establish an optimal inclusion cohort of both subjects and samples. Furthermore, I compared different per-allele association methods for identifying SNP/CpG associations and different analysis conditions (such as covariate data, statistical assumptions and confounding linkage effects) to ensure the accurate identification of independently-associated methQTLs.

I used this analytical framework to identify methQTLs across 333 subjects with matched genotype and methylation data. I identified 446,482 significant methylation QTLs, of which 13,195 were independently associating loci. There were 6,725 distinct SNPs which associated with methylation changes at 5,779 unique CpG sites. A higher proportion of breast tumour methQTLs were acting in cis (10,500) than in trans (2,695), although this ratio was more balanced relative to normal blood methQTLs. I then characterised the mechanisms underpinning cis-methQTLs and trans-methQTLs in breast tumours by analysing their enrichment in genomic elements which have functional importance. There was a depletion in tumour cis-methQTL-CpGs from CpG islands (both promoters and enhancers). Additionally, there was cis-methQTL-CpG enrichment in CTCF sites and depletion in bivalent promoters and repressed regions. This suggests that methylation change occurs in CpG island regulatory elements which are typically unmethylated and targets for hypermethylation in tumours, and therefore support a mechanism whereby instances of local hypermethylation may disrupt TF binding regions (such as CTCF sites) and polycomb-repressed bivalent promoters. In trans-methQTLs however, there was CpG enrichment in CGIs and promoters but depletion in enhancers. Meanwhile, CTCF site enrichment and bivalent promoter depletion were seen commonly across both cis- and trans-methQTL-CpGs. This suggests that methylation changes are primarily at distal promoters and supports the hypothesised mechanism that these are primarily cis-QTL mechanisms for the expression of transcription factors that subsequently programme longrange methylation effects. Overall, the concordance between the genomic profiles of methQTLs in both tumour and normal breast methylomes suggests that trans-methQTL mechanisms are more consistent compared with cis-methQTL mechanisms.

Taken together, these results show that germline genetic variation has a strong relationship with DNA methylation in breast cancer which associates with distinct methylome alterations from the normal blood methylome. Furthermore, this relationship may have a significant role in breast cancer biology by the mediation of transcriptional regulation through both local and long-range mechanisms.